BUG: DataFrame.rank does not return EA types when original type was an EADtype

[sharkipelago]

• closes BUG: DataFrame.rank does not return EA types when original type was an EADtype #52829
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.

I rewrote the method using _mgr.apply but still struggling to figure out why 2 of test_rank.py test cases are failing. Any help or tips are appreciated! They are as follows:

self = <pandas.tests.frame.methods.test_rank.TestRank object at 0x77da84fae490>

    def test_rank2(self):
        df = DataFrame([[1, 3, 2], [1, 2, 3]])
        expected = DataFrame([[1.0, 3.0, 2.0], [1, 2, 3]]) / 3.0
        result = df.rank(1, pct=True)
        tm.assert_frame_equal(result, expected)

        df = DataFrame([[1, 3, 2], [1, 2, 3]])
        expected = df.rank(0) / 2.0
        result = df.rank(0, pct=True)
        tm.assert_frame_equal(result, expected)

        df = DataFrame([["b", "c", "a"], ["a", "c", "b"]])
        expected = DataFrame([[2.0, 3.0, 1.0], [1, 3, 2]])
        result = df.rank(1, numeric_only=False)
>       tm.assert_frame_equal(result, expected)

pandas/tests/frame/methods/test_rank.py:84:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
pandas/_libs/testing.pyx:53: in pandas._libs.testing.assert_almost_equal
    cpdef assert_almost_equal(a, b,
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

>   raise_assert_detail(
E   AssertionError: DataFrame.iloc[:, 1] (column name="1") are different
E
E   DataFrame.iloc[:, 1] (column name="1") values are different (100.0 %)
E   [index]: [0, 1]
E   [left]:  [1.5, 1.5]
E   [right]: [3.0, 3.0]
E   At positional index 0, first diff: 1.5 != 3.0

pandas/_libs/testing.pyx:171: AssertionError

self = <pandas.tests.frame.methods.test_rank.TestRank object at 0x77da84faf290>
float_string_frame =                A         B         C         D  foo                   datetime       timedelta
foo_0   0.189053 -0.522...5.169587 1 days 00:00:01
foo_29 -0.967681  1.678419  0.765355  0.045808  bar 2025-08-25 12:57:35.169587 1 days 00:00:01

    def test_rank_mixed_frame(self, float_string_frame):
        float_string_frame["datetime"] = datetime.now()
        float_string_frame["timedelta"] = timedelta(days=1, seconds=1)

        float_string_frame.rank(numeric_only=False)
>       with pytest.raises(TypeError, match="not supported between instances of"):
E       Failed: DID NOT RAISE <class 'TypeError'>

[jbrockmendel]

In the axis=1 case you need to transpose the whole dataframe, not block-by-block

[sharkipelago]

Hello! Any other changes I should make to this?

[jbrockmendel]

what about 2D EAs? in particular DatetimeArray, TimedeltaArray

[sharkipelago]

I removed the raise axis != 0 in arrays/base.py because I thought it might be safe now with the other changes but not sure if there are other things to consider?

[jbrockmendel]

do we have cases that get here?

[sharkipelago]

Besides the two 2d array tests I just added , test_rank2() raises that Not Implemented Error when that line is there

=============================================== short test summary info ================================================
FAILED pandas/tests/frame/methods/test_rank.py::TestRank::test_rank2 - NotImplementedError
FAILED pandas/tests/frame/methods/test_rank.py::TestRank::test_2d_extension_array_datetime - NotImplementedError
FAILED pandas/tests/frame/methods/test_rank.py::TestRank::test_2d_extension_array_timedelta - NotImplementedError
============================================ 3 failed, 131 passed in 5.58s =============================================

[sharkipelago]

For the third "sub test" in test_rank2:

df = DataFrame([["b", "c", "a"], ["a", "c", "b"]])
expected = DataFrame([[2.0, 3.0, 1.0], [1, 3, 2]])
result = df.rank(1, numeric_only=False)

[jbrockmendel]

i think we can avoid transposing at the block level and only transpose at the dataframe level (L9317). then pass axis=blk_values.ndim-1 (which will require updating the EA method)

[jbrockmendel]

this transpose should also be avoidable

[jbrockmendel]

@sharkipelago can you address comments